If this is your first visit, be sure to
check out the FAQ by clicking the
link above. You may have to register
before you can post: click the register link above to proceed. To start viewing messages,
select the forum that you want to visit from the selection below.
prove this!... and i'm sure you "can't"
you only don't get a acceleration on lowend cards like hd4350 or hd4550.
you need at minimum a hd4650 to get a speed up compared to the cpu.
and this is your real secret point you claim linux users only use lowend hardware and on lowend hardware the hd4000 generation can not "accelerate"
anything above hd4650 is much faster than the "CPU".
(rolls eyes and gives up, wonders if it's too late to sign the petition)
there can not be a proof for "taste" "Personal Opinions"
but there can be a proof for a objective fact and OpenCL Performance on hd4000 hardware is not a taste or personal opinion its a objective fact .
and you claim that the objective fact the hd4000 openCL performance is "LOW" to "LOW" to get a acceleration.
this is a "Logical argument" now you can prove it right or "wrong"
you prefer to not prove it "right" and i proved this wrong with real world benchmarks.
i'm still waiting for a real work benchmark with the result that the openCL acceleration over a hd4870 card is slower than the CPU.
But its best we have, ok Intel is not that bad, too, but you dont get any information if you write in any forum or something like that (or am I wrong here?).
But Nvidia sucks extremely no free drivers... so I have to support this... But I hope that some stuff comes in next few months. The killer-feature is beside good power-management that works kind of ok in some cases, the gpu-decoding of videos.
Its just retarded to decode videos with the cpu in 2012, we have global warming, and I dont buy a laptop/... to make my battery empty in double the time possible.
I can wait very long on gaming speed, because gaming is dead (if you dont count wine) in Linux, you dont need much 3d speed anyway, at least most people. If Intel makes gpu encoded videos inkl. 1080p happen with its intel gpus + open driver and the hardware is not extremly more expensive I only will buy Intel stuff from this point, and also tell anybody who asks me about new hardware ( a few people ) that they should do so too.
I know only if you buy 10 trillion cpus/gpus from amd that is importent for amd, but dont forget that I dont speak only for myself, if people go to demonstrate its not only this few tausends that are voting for this or against this topic, there are much more people behind it that think in the same direction and do the same (as example in voting for the other parties).
I am very patiant, if you explain why it takes longer and tell me that the opensource driver is the focus, I can be pationed, but if its a thing that you do just for fun and dont take serios I think the last people who dont say you want a gpu under linux buy nvidia the "driver" (the blobshit) works better.
what did last year or this year important happen feature wise in the opensource ati driver? Maybe I am not very attentive but I would say nothing.
(rolls eyes and gives up, wonders if it's too late to sign the petition)
....... anyway bridgman, what is your take on this ?
in June 2011 Phil Rogers, AMD Fellow, said that Fusion processors – the ones that include CPU and GPU cores – would become CPU and GPU agnostic.
then AMD has followed up at this year's Fusion developer summit with the announcement of the formation of the Heterogeneous System Architecture (HSA) group , and then this week at the AMD Fusion Developer Summit 2012 (AFDS),
"AMD Senior Vice President and Chief Information Officer Mike Wolfe described AMD’s vision to advance computing security by enhancing AMD’s existing security technologies. This is expected to include developing a platform security processor using an ARM Cortex™-A5 CPU that features TrustZone technology, to monitor and help protect against malicious access to sensitive data and operations at the hardware level."
also given the Cortex-A5 has both the NEON SIMD and is using ARM's AMBA4 interconnect
"ARM Documentation set for AMBA protocol specifications and design tools, including ACE, ACE-Lite, AXI4, AXI4-Lite, AXI4-Stream, AXI3, ATB, AHB-Lite and APB protocols.
The AMBA protocol is an open standard, on-chip interconnect specification for the connection and management of functional blocks in a System-on-Chip (SoC). It enables development of multi-processor designs with large numbers of controllers and peripherals. AMBA promotes design re-use by defining a common backbone for SoC modules using specifications for ACE, AXI, AHB, APB and ATB."
will you be involved with the ARM AMD Linux documentation and managing the ARM Linux code that AMD writes, come to that will you and your Linux team now be joining the Linaro Initiative to get the AMD ARM code up to scratch ASAP along side all the other ARM vendors there.
and do You think AMD be using that NEON SIMD and the the low power ARM AMBA4 bus protocols etc http://www.arm.com/products/system-ip/amba/index.php
to interconnect your version of the Cortex-A5 to the rest of the AMD system APU's to reserve power or some other old x86 power hungry bus
anyone can explain to me how this will be good?? cuz i can't see any advantage only problems!!
* ""Security"" on hardware was never a good idea (Intel Vpro anyone??) to begin with an if security on hardware is updateable from software even worse idea.
* Low Power for some tasks!! mmm ok could be but how the hell the os is gonna manage pass from X86 ISA to ARM ISA, ofc not counting SSE/AVX to NEON on realtime?? i foresee a real mess here
* if AMD put some sort of hardware converter to keep the ARM ISA internal i only see massive time increase in wakeups and major performance scaling issues and prolly a fuck fest of possible hardware bugs from the switching
* if AMD is only interested in the security crap of trustzone, why in the freezing hell they use ARM and not port it to X86?? saddism??
* AMBA/ARM bandwith will be an issue? or hypertransport will downscale to SoC level before switching to ARM?
* how to debug crap when the ISA switch make a booboo??
* are they thinking in just handle ARM/X86 independently so you can multiarch ARM/ia32/ia64 in 1 CPU? it make any sense if you don't develop android apps?
this news has to be the most fucked up i heard this last decade LOL, bridgman memo your idiotic CEO and tell him to pour money in the CPU INTEGER performance and MEMORY controllers LOL and bulldozer could use an improvement in your AVX float performance too you know, i mean is competitive but X86/gpu performance is what sells CPU not some retarded "security" ARM crap superglued to the CPU or im missing something really uber important here?
Did I understand correctly the ARM core is just here to provide internal functionnalities to the graphics board (as the arm core is in some Marvell wifi chip) and it's not visible by the motherboard ?
still not the same since those are ARM microcontrollers and Wifi require very little performance but i just can imagine and ARM cortex A5 could keep up with a PCIe while doing this security crap and feed the GPU too ofc while the x86 ISA is translated to ARM, there must be something else